Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Abstract Cyclists are overrepresented among motor vehicle crash fatalities. Detailed information regarding
common cyclist crash scenarios and relevant crash factors is crucial to the development of cyclist detection
warning and crash avoidance systems that could prevent these crashes and fatalities. Motor vehicle‐cyclist crash
data from federally maintained national databases were used to identify common and fatal crash scenarios
between cyclists and motor vehicles. The most common fatal crash modes involved the motor vehicle‐cyclist
movement combinations straight‐in line, straight‐crossing, and straight‐against. The most common crash modes
involved the movement combinations straight‐crossing, turning‐crossing, and turning‐in line. Crashes that
occurred in non‐daylight conditions and on roads with speed limits of 40 mi/h and greater contributed to the
greatest percentage of fatalities. Cyclist detection systems that function at high speeds and in both daylight and
non‐daylight conditions offer the greatest potential benefit. Effective cyclist detection systems designed to
function in scenarios like the three common fatal crash modes and two additional most common crash modes
could help mitigate or prevent up to 47% of crashes, 48% of injuries, and 54% of fatalities, potentially saving up
to 363 lives annually.
Keywords Bicyclist, cyclist detection, crash avoidance, vulnerable road users
I. INTRODUCTION
In recent years, motor vehicle‐cyclist crash fatalities have begun to increase in the United States, following
decades of decline since the all‐time high of 1,003 cyclist deaths in 1975 [1]. The 682 cyclist deaths in 2011
represented a 10% increase from 2010, and 2012 saw an additional 6% increase to 722 cyclist fatalities [2].
Although the approximately 700 cyclists who are killed each year in crashes with motor vehicles represent a
small fraction of the total annual motor vehicle crash deaths, cyclists are disproportionately likely to be fatally
injured. Barely 1% of person road trips are attributed to cyclists, but these riders account for 2% of roadway
fatalities [3]. Cyclists are required to follow the same regulations and traffic patterns as motor vehicles with
little to none of the protection afforded to motor vehicle occupants. Cyclist protection becomes increasingly
important as urban areas make efforts to reduce environmental and traffic concerns by promoting cycling as
primary means of transportation.
Historically, efforts to improve cyclist safety have emphasized helmet use and adherence to riding regulations.
Infrastructure can also have a substantial effect on cyclist safety. Studies have shown that cyclist‐specific
facilities like roundabouts and bike lanes engineered with an emphasis on cyclist safety reduce crashes and
injuries to cyclists [4]. Other improvements including adequate street lighting, paved surfaces, and low‐angled
road grades are beneficial to cyclist safety [4].
While improvements to the road environment reduce the risk to cyclists, changes to the vehicles navigating
the roadways can also help. The percentage of the passenger vehicle fleet with new collision warning and crash
avoidance technologies is growing relatively quickly. Crash warning and avoidance systems already make use of
different types of sensor technologies including camera‐based systems, night vision technology, radar, and
LIDAR systems [5]. Each of these sensor technologies has its own strengths and weaknesses. For example,
camera systems require sufficient ambient light to perform well, while radar and LIDAR systems can be
adversely affected by even small amounts of precipitation. Each type of system has its own effective range,
which has great bearing on warning and intervention timing. The development of effective cyclist detection
systems must take into account the different strengths and limitations of each sensor technology as they apply
Anna MacAlister is a research engineer and David S. Zuby is the Executive Vice President and Chief Research Officer for the Insurance Institute for Highway Safety 1005 North Glebe Road, Arlington, VA 22201, United States. (Corresponding author. tel.: 1 434 985 4600; fax: 1 434 985 2202; e‐mail: [email protected] ).
Cyclist Crash Scenarios and Factors Relevant to the Design of Cyclist Detection Systems
Anna MacAlister, David S. Zuby
IRC-15-50 IRCOBI Conference 2015
- 373 -
to different crash scenarios.
The European New Car Assessment Programme (Euro NCAP) has recently taken steps towards reducing the
number of pedestrian and cyclist deaths in Europe [6]. Vulnerable road users (VRU), a category that includes
pedestrians and cyclists, account for approximately half of all deaths on EU‐27 roadways [6]. For this reason,
Euro NCAP plans to develop protocols for testing pedestrian and cyclist detection and adopt testing procedures
for autonomous emergency braking VRU systems no later than 2018 [6]. A recent study of German cyclist
crashes found that the three most common motor vehicle‐cyclist crash scenarios all occur when a cyclist crosses
the path of a motor vehicle, either from the right while the motor vehicle travels straight or executes a right
hand turn, or from the left as the motor vehicle travels straight [7]. While these data will certainly influence the
development of cyclist detection systems for the European market, an understanding of the most common and
most severe cyclist crash modes in the United States is necessary to the development of broadly applicable and
effective cyclist detection systems. Given the different infrastructures, driving patterns, vehicle types, and cyclist
cultures of Europe and the United States, it is not necessarily a valid assumption that the technologies that will
produce the largest effect in one market will be equally beneficial in the other. The design and implementation
of cyclist detection, collision warning, and crash avoidance systems must be guided by relevant data regarding
the most common, fatal, and preventable crash modes.
II. METHODS
Motor vehicle‐cyclist crash data from federally maintained national databases were used to identify common
and fatal crash scenarios between cyclists and motor vehicles, similar to a 2011 study conducted by the
Insurance Institute for Highway Safety evaluating pedestrian crash scenarios [8]. Crash scenarios were defined
by seven factors that describe the paths of the cyclist and motor vehicle, as well as external factors related to
the crash with implications for collision warning and avoidance systems. The selected factors may affect sensor
type selection and the ability of detection systems to accurately identify cyclists in a time frame that allows for
driver or automated intervention to prevent a collision from occurring. Information describing the motion of a
cyclist relative to the vehicle can help identify sensor ranges and fields of view, as well as help define the logic
needed to develop algorithms that estimate crash risk. Obstruction of driver view is used as a proxy for the
possibility that detection sensors may not have a clear view of a cyclist. Light condition provides information
regarding the ability of light‐dependent sensors in detecting cyclists. Speed limit, used as a proxy for vehicle
speed, affects the necessary distance range of sensing technologies, as well as the time to collision and timing of
warnings and interventions. Cyclist age is used as a proxy for cyclist size, an important variable in detection
algorithms. Weather conditions, particularly precipitation and other particulate (blowing sand, etc.), affect the
ability of the vehicle to brake, either autonomously or through driver intervention, as well as the time needed to
apply an intervention. Driver braking affects the design of systems intended to autonomously brake or support
and enhance driver avoidance maneuvers.
Data were extracted from two national crash databases maintained by the National Highway Traffic Safety
Administration (NHTSA) in the United States. The National Automotive Sampling System General Estimates
System (NASS GES) is a nationally representative sample of police‐reported motor vehicle crashes that occur on
public roads. The cases are weighted by the inverse of their selection probability to provide a sample that is
representative of approximately 6 million annual motor vehicle crashes nationwide. While NASS GES weights
can be used to estimate fatal crash counts, information about these crashes was extracted from the Fatality
Analysis Reporting System (FARS). FARS is a census of motor vehicle crashes in which at least one occupant or
other involved person was fatally injured and died within 30 days of the crash.
All person‐level records for cyclists in the 2008‐2012 NASS GES and FARS files were extracted and combined
with their corresponding crash‐level records. For crashes involving a single motor vehicle, the cyclist and crash
data were again merged with the corresponding driver and vehicle records. Crashes involving more than one
motor vehicle were not included in the study to restrict the analysis to primary motor vehicle‐cyclist crashes and
eliminate the confounding factors that could be introduced when a cyclist crash occurred as a result of a
previous vehicle‐vehicle crash. NASS GES cases were weighted by their case weights to produce national
estimates of motor vehicle‐cyclist crash incidences.
Crashes involving the front of the vehicle were further analyzed to isolate the effects of crash factors on
scenarios that could be preventable with a cyclist detection algorithm. Cyclist crashes were classified according
IRC-15-50 IRCOBI Conference 2015
- 374 -
to the seven primary accident factors (Table 1).
TABLE 1 CLASSIFICATION OF CYCLIST AND VEHICLE MOVEMENT, DRIVER VIEW OBSTRUCTION, LIGHT CONDITIONS, INCLEMENT WEATHER,
CYCLIST AGE, SPEED LIMIT, AND VEHICLE BRAKING USING NASS GES AND FARS RECORDS, 2008‐2012
NASS GES FARS
Cyclist movement
Cyclist in‐line with traffic PED_ACC = 13, 15‐17, 22‐24,
27, 35, 39, 61, 62
OR
MPR_ACT = 5
LOCATION = 11‐13, 15, 16
AND
P_CF1‐3 ≠ 3, 47‐50
OR
MPR_ACT = 5
Cyclist against traffic PED_ACC = 26, 30
OR
MPR_ACT = 6
P_CF1‐3 = 49, 50
OR
MPR_ACT = 6
Cyclist crossing traffic PED_ACC = 1, 2, 4‐10, 12, 18, 19, 21,
25, 31‐34, 48, 49, 50, 55, 60, 90
OR
MPR_ACT = 3
P_CF1‐3 = 3, 47, 48
OR
MPR_ACT = 3
Vehicle movement
Vehicle traveling straight P_CRASH1 = 1‐4, 6, 14‐16 VEH_MAN = 1‐5, 9, 16, 17
OR
P_CRASH1 = 1‐4, 6, 14‐16
Vehicle turning P_CRASH1 = 10‐12 VEH_MAN = 10‐14
OR
P_CRASH1 = 10‐12
Driver view obstruction reported VIS_OBSC = 1‐15, 97, 98
OR
MVISOBSC = 1‐14, 97, 98
DR_CF1‐4 = 61‐76
OR
MVISOBSC = 1‐14, 97, 98
Non‐daylight conditions LGHT_CON = 2‐5
OR
LGT_COND = 2‐6
LIGHT = 2‐6
OR
LGT_COND = 2‐6
Inclement weather WEATHER = 2‐8, 10, 11
OR
WEATHER1‐2 = 2‐8, 10, 11
WEATHER = 2‐8, 10, 11
OR
WEATHER1‐2 = 2‐8, 10, 11
Cyclist age 12 years or younger AGE = 0‐12 AGE = 0‐12
Speed limit less than 40 mi/h SPD_LIM = 1‐39
OR
VSPD_LIM = 1‐39
SP_LIMIT = 1‐39
OR
VSPD_LIM = 1‐39
Vehicle braking reported IMPACT1 = 1, 11, 12, 21, 31, 32
AND
P_CRASH3 = 2‐5, 8, 9
IMPACT1 = 1, 11, 12
AND
P_CRASH3 = 2‐5, 8, 9
Cyclist movement was classified as in‐line with traffic, against traffic, or crossing traffic. Motor vehicle
movement prior to the crash was classified as traveling straight or turning. Six movement combinations
obtained using the three cyclist movement types and two vehicle movement types were used to classify crash
IRC-15-50 IRCOBI Conference 2015
- 375 -
scenarios (Figure 1). Cases that occurred in non‐daylight conditions or during inclement weather, as well as
cases in which the driver’s view was obstructed, were identified using variables coded within each database
(Table1).
Fig. 1 Depictions of motor vehicle and cyclist movement combinations for crash scenarios (a) vehicle moving
straight and cyclist traveling in line with traffic, (b) vehicle moving straight and cyclist crossing traffic, (c) vehicle
moving straight and cyclist moving against traffic, (d) vehicle turning and cyclist moving in line with traffic, (e)
vehicle turning and cyclist crossing traffic, and (f) vehicle turning and cyclist moving against traffic.
III. RESULTS
From 2008 to 2012, there were 7,835 cyclist records in NASS GES, representing approximately 275,000 cyclists
with case weights applied and rounded to the nearest thousand. During the same time frame, 3,367 cyclist
fatalities were documented in FARS. During this 5‐year period, an average of 55,000 cyclists were involved in
police‐reported motor vehicle crashes each year, and an average of 673 cyclists per year sustained fatal injuries
in motor vehicle crashes. Cyclist crashes involving a single motor vehicle accounted for 99% of all cyclist motor
vehicle crashes and 95% of cyclist fatalities for a total of 274,000 cyclist‐motor vehicle crashes and 3,201 deaths
during the 5‐year study period. The distribution of cyclist crashes, injuries, and fatalities by various vehicle,
environmental, and cyclist characteristics are shown in Table 2.
Front impacts to single passenger vehicles
During the 5‐year study period, cyclist crashes involving the front of a passenger vehicle accounted for 63%
(172,000) of all single motor vehicle cyclist crashes and 74% (2,363) of fatal cyclist crashes. The distribution of
cyclist crashes, injuries, and fatalities by vehicle, environmental, and cyclist characteristics for crashes involving
the fronts of passenger vehicles are shown in Table 3. Sixty‐four percent of passenger vehicle‐cyclist crashes
with the front of the vehicle occurred on roads with speed limits of less than 40 mi/h, while only 35% of fatal
crashes occurred on roads with the same speed limits. Twenty‐three percent of cyclist crashes and 50% of fatal
crashes occurred during non‐daylight conditions. Inclement weather was present in 11% of all cyclist crashes
and 13% of fatal cyclist crashes. Children 12 and younger accounted for 12% of cyclist crashes but only 7% of
cyclist fatalities.
IRC-15-50 IRCOBI Conference 2015
- 376 -
TABLE 2 PERCENT DISTRIBUTION OF CYCLISTS IN SINGLE‐VEHICLE CRASHES BY
VEHICLE, ENVIRONMENT, AND CYCLIST CHARACTERISTICS, 2008‐2012
All cyclists
N=274,000
Injured cyclists
N=147,000
Fatally injured cyclists
N=3201
Vehicle type
Passenger vehicle 97.6 97.3 83.0
Heavy truck or bus 1.7 2.1 11.1
Motorcycle 0.7 0.6 0.9
Other or unknown 0.0 0.6 5.1
Point of initial impact on vehicle
Front 63.8 64.3 83.6
Right 22.9 23.0 7.0
Left 9.0 9.2 2.8
Rear 4.1 3.4 2.1
Other or unknown 0.3 0.1 4.5
Speed limit
No limit 6.7 7.1 0.0
<30 30.3 30.6 10.6
30‐39 32.9 34.1 27.1
40‐49 11.8 13.2 30.6
>49 3.8 3.1 26.6
Unknown 14.6 11.9 5.0
Light condition
Daylight or unknown 77.9 79.0 53.2
Dark 3.6 4.1 20.0
Dark but lighted 13.6 13.2 21.3
Dawn/dusk 4.8 3.8 5.5
Inclement weather 11.4 11.3 13.1
Cyclist age
<6 0.7 0.8 0.9
6‐12 10.5 11.3 5.6
13‐17 16.0 15.1 7.0
18‐64 68.4 68.5 74.4
>64 4.4 4.2 12.1
Unknown 0.0 0.0 0.0
Male cyclists 79.1 79.0 86.8
Cyclist location
Intersection 35.3 33.5 32.1
Roadway, non‐intersection 31.7 31.6 60.9
Dedicated bike lane* 1.4 1.7 1.4
Unknown 31.6 33.2 5.6
Cyclist wearing helmet 16.5 18.4 12.8
*Prior to 2010, information on dedicated bike lanes was not available in the NASS GES and FARS databases.
IRC-15-50 IRCOBI Conference 2015
- 377 -
TABLE 3 PERCENT DISTRIBUTION OF CYCLISTS IN SINGLE‐VEHICLE CRASHES INVOLVING FRONTS OF
PASSENGER VEHICLES BY VEHICLE, ENVIRONMENT, AND CYCLIST CHARACTERISTICS, 2008‐2012
All cyclists
N=172,000
Injured cyclists
N=93,000
Fatally injured cyclists
N=2363
Speed limit (mi/h)
No limit 8.1 8.1 0.0
<30 30.4 31.7 9.4
30‐39 33.1 32.7 25.1
40‐49 11.3 13.8 32.8
>49 3.3 2.6 29.1
Unknown 13.8 11.2 3.6
Light condition
Daylight or unknown 76.5 76.7 49.9
Dark 3.9 4.7 22.4
Dark but lighted 14.7 14.7 22.6
Dawn/dusk 4.8 3.9 5.2
Inclement weather 11.4 11.6 13.2
Cyclist age
<6 0.9 1.0 0.6
6‐12 11.3 12.2 5.9
13‐17 17.7 15.9 7.5
18‐64 65.6 66.4 74.0
>64 4.5 4.5 11.9
Unknown 0.0 0.0 0.0
Male cyclists 78.2 78.2 87.5
Cyclist location
Intersection 35.1 33.4 30.3
Roadway, non‐intersection 27.8 26.9 63.1
Dedicated bike lane* 0.8 1.0 1.5
Unknown 36.3 38.6 5.2
Cyclist wearing helmet 13.6 14.8 12.8
Vehicle movement
Traveling straight 50.1 50.6 93.3
Turning 44.0 44.9 6.1
Other or Unknown 5.9 4.4 0.6
Cyclist movement
Crossing traffic 54.2 54.1 24.1
Moving in‐line with traffic 20.8 23.2 47.0
Moving against traffic 6.7 5.5 6.7
Other or unknown 18.3 17.2 22.3
Driver view obstruction reported 9.1 10.2 6.0
Vehicle braking reported 6.3 7.0 10.4
*Prior to 2010, information on dedicated bike lanes was not available in the NASS GES and FARS databases.
IRC-15-50 IRCOBI Conference 2015
- 378 -
Seventy‐eight percent of cyclists involved in crashes and 88% of cyclists fatally injured in crashes were male.
The majority of fatal cyclist crashes (63%) occurred on roadways outside of intersections, where only 28% of all
crashes occurred. Fourteen percent of cyclists involved in a crash were reported to be wearing a helmet, as
were 13% of fatally injured cyclists. About half of all vehicle‐cyclist crashes involved a vehicle traveling straight.
The same vehicle movement was present in 93% of fatal vehicle‐cyclist crashes. While half of all cyclist crashes
occurred when the cyclist was crossing traffic, less than a quarter of all fatalities occurred when cyclists crossed
traffic. In 9% of vehicle‐cyclist crashes and 6% of fatal crashes, the driver’s view was reportedly obstructed.
Vehicle braking was reported in 6% of all cyclist crashes and 10% of fatal crashes.
Crash modes and movement combinations
Cyclist and vehicle movement combinations are presented in Table 4. The three crash modes leading to the
most fatalities involved the following motor vehicle‐cyclist movement combinations: straight‐in‐line (1072
fatalities), straight‐crossing (530 fatalities), and straight‐against (139 fatalities). The most common crash mode
involved the same vehicle‐cyclist movement combination as the second most fatal crash mode, straight‐
crossing, accounting for 50,000 crashes. The next most common crash modes involved the following vehicle‐
cyclist movement combinations: turning‐crossing (39,000 crashes) and turning‐in‐line (18,000 crashes). Crashes
that produced injuries, but not fatalities, followed the same vehicle‐cyclist movement patterns as all cyclist
crash involvements.
Crash factors
Summaries of the influence of crash factors for each of the three crash scenarios that result in the greatest
numbers of fatalities are presented in Tables 5 and 6. Sixteen percent of involved cyclists were children ages 12
and younger, but only 6% of fatal crashes involved children of this age group. Twenty‐nine percent of cyclists
involved in the three most common fatal crash scenarios crashed in nighttime conditions, whereas more than
half of fatal crashes were at night. Inclement weather was a factor in 13% of all crashes, with 14% of all fatal
crashes occurring in these scenarios. Driver view obstruction was reported in 9% of the three most frequent
fatal crash scenarios and 6% of fatal crashes included these types. Two‐thirds (67%) of cyclist crashes of the
three common fatal crash modes occurred on roads with speed limits of less than 40 mi/h, while less than one‐
third (30%) of fatal crashes occurred on these roads. Sixty‐six percent of fatal crashes in these modes occurred
on roads with speed limits of less than 50 mi/h. Lastly, vehicle braking was reported in 8% of crashes and 11% of
fatal crashes.
Summaries of the influence of crash factors for each of the three most common crash scenarios are presented
in Tables 7 and 8. Thirteen percent of cyclists involved in the three most common motion scenarios were
children ages 12 and younger, but only 8% of fatal crashes in these scenarios involved children of this age group.
Twenty‐four percent of the crashes in the most common scenarios occurred at night, as did approximately half
of fatal crashes. Inclement weather was a factor in 11% of all common crashes and 15% of all fatal common
crashes. Driver view obstruction was reported in 8% of the most common crash scenarios and 7% of fatal
crashes occurring in these three scenarios. More than two‐thirds (69%) of common cyclist crashes and 42% of
common crashes leading to cyclist fatalities occurred on roads with speed limits of less than 40 mi/h. Finally,
vehicle braking was reported in 7% of crashes and 16% of fatal crashes.
Some crash scenarios involved more than one crash factor. Table 9 shows hierarchical distributions of the
number of crashes and fatalities that could be addressed by cyclist detection, collision warning, and crash
avoidance systems designed to take into account each of the crash factors for the three most common fatal
crash scenarios and two additional common crash scenarios. During the 5‐year study period, 54% of all fatal
cyclist crashes with the front of a single passenger vehicle occurred in the three most common fatal crash
modes or the remaining two most common modes. These same five crash modes accounted for approximately
74% of all cyclist crashes with the fronts of single passenger vehicles.
IRC-15-50 IRCOBI Conference 2015
- 379 -
TABLE 4
VEH
ICLE AND CYC
LIST M
OVEM
ENT COMBINATIONS FO
R CYC
LIST IN
VOLVEM
ENTS, INJURIES, AND
DEA
THS IN SINGLE‐VEH
ICLE CRASH
ES W
ITH FRONTS OF PASSEN
GER
VEH
ICLES, 2008‐2012
Cyclist movemen
t
Crossing traffic
In‐line
Against
Other/unknown
Total
Veh
icle m
ovemen
t All
Injured
Fatal
All
Injured
Fatal
All
Injured
Fatal
All
Injured
Fatal
All
Injured Fatal
Straight
50000
27000
530
16000
10000
1072
5000
2000
139
15000
7000
460
86000
47000 2205
Turning
39000
21000
38
18000
11000
30
6000
2000
18
13000
7000
57
76000
42000
143
Other/unknown
3000
2000
12000
1000
81000
01
4000
1000
510000
4000
15
Total
93000
50000
569
36000
22000
1110
12000
5000
158
31000
16000
522
172000
93000 2363
TA
BLE
5
CYC
LIST IN
VOLVEM
ENTS AND FATA
LITIES IN
SINGLE‐VEH
ICLE CRASH
ES W
ITH FRONTS OF PASSEN
GER
VEH
ICLES BY CYC
LIST AND
ENVIRONMEN
T CRASH
FACTO
RS FO
R THREE M
OST COMMON FATA
L CRASH
SCEN
ARIOS WITH GREA
TEST NUMBER
OF FA
TALITIES, 2008‐2012
To
tals
Children 12 and younger
Non‐daylight conditions
Inclem
ent weather
Veh
icle‐cyclist
Involvem
ents
Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
movemen
t N
NN (%)
N (%)
N (%)
N (%)
N (%)
N (%)
Straight‐in line
16000
1072
1000 (7.7)
37 (3.5)
6000 (34.7)573 (53.5)
2000 (15.0)145 (13.5)
Straight‐crossing
50000
530
10000 (19.7)
46 (8.7)
14000 (27.9)276 (52.1)
6000 (11.3)
80 (15.1)
Straight‐against
5000
139
500 (9.8)
12 (8.6)
1000 (23.4)
75 (54.0)
1000 (17.8)
21 (15.1)
Total
71000
1741
12000 (16.3)
95 (5.5)
21000 (29.2)924 (53.1)
9000 (12.6)246 (14.1)
TA
BLE 6
CYC
LIST IN
VOLVEM
ENTS AND FATA
LITIES IN
SINGLE‐VEH
ICLE CRASH
ES W
ITH FRONTS OF PASSEN
GER
VEH
ICLES BY
VEH
ICLE CRASH
FACTO
RS FO
R THREE M
OST COMMON FATA
L CRASH
SCEN
ARIOS WITH GREA
TEST NUMBER
OF FA
TALITIES, 2008‐2012
To
tals
Driver view
obstruction
Speed limit <40 m
i/h
Speed limit <50 m
i/h
Veh
icle braking
Veh
icle‐cyclist
Involvem
ents Fatalities
Involvem
ents
Fatalities
Involvem
ents Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
movemen
t N
NN (%)
N (%)
N (%)
N (%)
N (%)
N (%)
N (%)
N (%)
Straight‐in line
16000
1072
1000 (6.8)
60 (5.6)
9000 (57.3) 277 (25.8)
13000 (77.0)
633 (59.0)
1000 (6.0)
77 (7.2)
Straight‐crossing
50000
530
4000 (8.5)
36 (6.8)
36000 (71.0) 202 (38.1)
42000 (83.5)
423 (79.8)
4000 (7.7)
92 (17.4)
Straight‐against
5000
139
1000 (13.1)
5 (3.6)
2000 (51.9)
44 (31.7)
3000 (60.0)
84 (60.4)
1000 (16.8)
21 (15.1)
Total
71000
1741
6000 (8.7)101 (5.8)
48000 (66.6) 523 (30.0)
57000 (80.5)1140 (65.5)
6000 (7.9)190 (10.9)
IRC-15-50 IRCOBI Conference 2015
- 380 -
TABLE
7
CYC
LIST IN
VOLVEM
ENTS AND FATA
LITIES IN
SINGLE‐VEH
ICLE CRASH
ES W
ITH FRONTS OF PASSEN
GER
VEH
ICLES BY
CYC
LIST AND ENVIRONMEN
T CRASH
FACTO
RS FO
R THREE M
OST COMMON CRASH
SCEN
ARIOS, 2008‐2012
To
tals
Children 12 and younger
Non‐daylight conditions
Inclem
ent weather
Veh
icle‐cyclist
Involvem
ents
Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
movemen
t N
NN (%)
N (%)
N (%)
N (%)
N (%)
N (%)
Straight‐crossing
50000
530
10000 (19.7)
46(8.7)
14000 (27.9)
276 (52.1)
6000 (11.3)
80 (15.1)
Turning‐crossing
39000
38
3000 (6.9)
3 (7.9)
7000 (18.9)
12 (31.6)
4000 (9.9)
3 (7.9)
Turning‐in line
18000
30
1000 (5.1)
1 (3.3)
5000 (26.0)
7 (23.3)
2000 (9.9)
4 (13.8)
Total
107000
598
14000 (13.1)
50 (8.4)
26000 (24.3)
295 (49.3)
12000 (11.2)
87 (14.5)
TA
BLE 8
CYC
LIST IN
VOLVEM
ENTS AND FATA
LITIES IN
SINGLE‐VEH
ICLE CRASH
ES W
ITH FRONTS OF PASSEN
GER
VEH
ICLES BY
VEH
ICLE CRASH
FACTO
RS FO
R THREE M
OST COMMON CRASH
SCEN
ARIOS, 2008‐2012
To
tals
Driver view
obstruction
Speed limit <40 m
i/h
Speed limit <50 m
i/h
Veh
icle braking
Veh
icle‐cyclist
Involvem
ents Fatalities
Involvem
ents
Fatalities
Involvem
ents Fatalities
Involvem
ents
Fatalities
Involvem
ents
Fatalities
movemen
t N
NN (%)
N (%)
N (%)
N (%)
N (%)
N (%)
N (%)
N (%)
Straight‐crossing
50000
530
4000 (8.5)
36(6.8)
36000 (71.0) 202 (38.1)
42000 (83.5)
423 (79.8)
4000 (7.7)
92 (17.4)
Turning‐crossing
39000
38
2000 (6.1)
3 (5.3)
27000 (68.9)
31 (81.6)
31000 (77.7)
34 (89.5)
2000 (3.8)
2 (5.3)
Turning‐in line
18000
16
2000 (12.4)
6 (20.0)
11000 (64.2)
17 (56.7)
14000 (77.1)
23 (76.7)
1000 (4.0)
1 (3.3)
Total
107000
598
8000 (7.5)
44 (7.4)
74000 (69.2) 250 (41.8)
87000 (81.3)
480 (80.3)
7000 (6.5)
95 (15.9)
IRC-15-50 IRCOBI Conference 2015
- 381 -
TABLE 9 CYCLIST INVOLVEMENTS, INJURIES, AND FATALITIES IN SINGLE‐VEHICLE CRASHES BY THREE MOST COMMON
FATAL AND TWO ADDITIONAL MOST COMMON CRASH SCENARIOS THAT COULD BE ADDRESSED IF CYCLIST
DETECTION SYSTEMS WERE ABLE TO FUNCTION IN NON‐DAYLIGHT CONDITIONS, AT SPEEDS GREATER THAN 40 MI/H, IN INCLEMENT WEATHER, AND WITH DRIVER VIEW OBSTRUCTION, 2008‐2012
Cyclists involved
Cyclists injured
Cyclist fatalities
Vehicle traveling straight, cyclist moving in line
Crashes addressed by base system
Speed limit <40 mi/h, daylight, clear weather, no view obstruction 5000 3000 112
Potential crashes addressed
Non‐daylight conditions and speed limit ≥40 mi/h 8000 6000 769
Inclement weather 2000 1000 131
Driver view obstruction 1000 1000 60
Subtotal for scenario 16000 10000 1072
Vehicle traveling straight, cyclist crossing
Crashes addressed by base system
Speed limit <40 mi/h, daylight, clear weather, no view obstruction 22000 12000 94
Potential crashes addressed
Non‐daylight conditions and speed limit ≥40 mi/h 19000 10000 327
Inclement weather 5000 2000 73
Driver view obstruction 4000 3000 36
Subtotal for scenario 50000 27000 530
Vehicle traveling straight, cyclist moving against
Crashes addressed by base system
Speed limit <40 mi/h, daylight, clear weather, no view obstruction 1000 1000 20
Potential crashes addressed
Non‐daylight conditions and speed limit ≥40 mi/h 2000 1000 94
Inclement weather 1000 1000 20
Driver view obstruction 1000 0 5
Subtotal for scenario 5000 2000 139
Vehicle turning, cyclist crossing
Crashes addressed by base system
Speed limit <40 mi/h, daylight, clear weather, no view obstruction 19000 10000 17
Potential crashes addressed
Non‐daylight conditions and speed limit ≥40 mi/h 15000 8000 16
Inclement weather 4000 2000 3
Driver view obstruction 2000 2000 2
Subtotal for scenario 39000 21000 38
Vehicle turning, cyclist traveling in line
Crashes addressed by base system
Speed limit <40 mi/h, daylight, clear weather, no view obstruction 6000 4000 9
Potential crashes addressed
Non‐daylight conditions and speed limit ≥40 mi/h 7000 4000 11
Inclement weather 2000 1000 4
Driver view obstruction 2000 1000 6
Subtotal for scenario 18000 11000 30
IRC-15-50 IRCOBI Conference 2015
- 382 -
IV. DISCUSSION
Crashes in which the vehicle was traveling straight and the cyclist was moving in line with traffic were found
to result in the greatest number of cyclist fatalities, followed by crashes in which the vehicle was traveling
straight and the cyclist was either crossing traffic or moving against traffic, respectively (Table 4). These three
crash modes alone account for 74% of cyclist fatalities in crashes to the fronts of passenger vehicles. Including
the additional 460 cyclists with unknown movement patterns who were fatally injured by straight‐moving
vehicles, more than 93% of cyclist fatalities in crashes to the fronts of passenger vehicles occurred when the
vehicle was traveling straight. The most common crashes followed a different pattern. Although the most
common crash mode, in which the vehicle was traveling straight and the cyclist was crossing traffic, also was the
second most common fatal crash scenario, the remaining two common crash modes were not among the most
common fatal scenarios. The second and third most common crash modes involved the vehicle turning and the
cyclist either crossing traffic or moving in line with traffic, respectively. The difference in crash patterns leading
to the greatest number of fatalities and those leading to the greatest number of crashes means that cyclist
detection systems must focus on a variety of crash scenarios. Additionally, crash factors affect the numbers of
fatalities and the total number of crashes differently. For example, while the majority of crashes occur at vehicle
speeds less than 40 mi/h, only one‐third of fatalities occur below 40 mi/h (Table 3). To prevent the majority of
crashes and fatalities, cyclist detection systems must function at vehicle speeds up to and greater than 50 mi/h.
Cyclist detection systems that function at high speeds and in both daylight and non‐daylight conditions have the
greatest potential to prevent cyclist fatalities. Existing forward collision prevention systems may have the
potential to address the most common fatal cyclist crash scenario, straight‐in line, with only minor
modifications. These systems already have been demonstrated to be effective at reducing the frequency of
property damage and bodily injury liability insurance claims [9]. A more recent study indicates that low‐speed
(≤50 km/h) autonomous emergency braking systems reduce real‐world rear impact crashes by 38% [10].
Extending the function of these systems to include the capability to identify cyclists may be possible with the
sensing technology already deployed in many front crash prevention systems. This one crash scenario accounts
for 6% of all cyclist crashes and 32% of fatal cyclist crashes (Table 9).
Although impacts to the front of the vehicle make up the largest portion of cyclist crashes and fatalities,
there is a non‐trivial number of crashes and fatalities that occur to the right side of the vehicle (Table 2). A
detailed analysis of crashes to the sides of vehicles could guide the development of new technologies designed
to prevent or mitigate another significant portion of cyclist crashes and fatalities.
While this study provides guidance for the development of cyclist detection and collision warning and
avoidance systems, it is subject to limitations. First, the data used in this study were obtained from two
databases, NASS GES and FARS, and are therefore subject to the limitations of the databases. Any errors made
during the initial coding of each case cannot be determined and are present in the data set. Second,
discrepancies exist between the two databases. For crashes occurring in 2010 and later, the NASS GES and FARS
databases use the same variables and field codes. For crashes occurring prior to 2010, each database follows a
unique coding procedure and variables in one dataset do not necessarily correspond directly to variables in the
other. Third, the variables coded in NASS GES and FARS do not provide a comprehensive view of the crash. For
example, the variable driver view obstruction can take on many values describing what blocked the view of the
driver, but the field does not provide information regarding to what degree the driver’s view was obstructed or
for what period of time leading up to the crash. This type of detailed information would be beneficial to the
development of cyclist detection systems. Finally, one fundamental limitation of all cyclist data in the NASS GES
and FARS databases is the way in which cyclist movement is coded. In both databases, before and after
standardization, cyclists and pedestrians are treated collectively as “non‐motorist” person types. While there
are some concessions made to account for their differences, like the bicycle helmet usage field, their movement
patterns are characterized similarly. Codes for non‐motorist movement are primarily derived from pedestrian
movement styles, rather than the movement patterns of cyclists. More detailed information regarding crash
scenarios could be attained if the NASS GES and FARS databases coded cyclist movement in a manner similar to
vehicle movement, reflecting the movement patterns cyclists are legally obligated to obey.
IRC-15-50 IRCOBI Conference 2015
- 383 -
V. CONCLUSIONS
The majority of all crashes involving a cyclist and motor vehicle occur to the front of passenger vehicles. Crashes in which the cyclist crosses the path of traffic are the most common in the United States, while the most common fatal crash mode involves a motor vehicle striking a cyclist moving along with traffic. Changes to traffic infrastructure and cyclist safety culture are vital in the effort to reduce these overrepresented traffic fatalities, but vehicle safety systems including cyclist detection and collision warning and avoidance also have the potential to greatly improve the safety of roadways for cyclists. For these types of systems to have the greatest possible benefit, it is important that they are designed with the most common and most severe crash scenarios in mind. Cyclist detection algorithms coupled with collision warning and crash avoidance systems designed with the three most common fatal scenarios and factors in mind could help mitigate or prevent up to 26% of crashes and cyclist injuries and 52% of fatalities. Systems designed with the remaining two most common crash modes in mind have the potential to mitigate or prevent up to an additional 21% of crashes, 22% of cyclist injuries, and 2% of cyclist fatalities, affecting a total of up to 47% of cyclist crashes, 48% of cyclist injuries, and 54% of cyclist fatalities.
VI. REFERENCES
[1] National Highway Traffic Safety Administration. 2012 bicyclists and other cyclists traffic safety fact sheet.
Report no. DOT HS‐812‐018. Washington, D.C.: U.S. Department of Transportation; 2014.
[2] Insurance Institute for Highway Safety. Pedestrian and bicyclists fatality facts. Insurance Institute for
Highway Safety, Arlington VA, 2014.
[3] Pucher J, Buehler R, Merom D, Bauman A. Walking and cycling in the United States, 2001‐2009: evidence
from the National Household Travel Surveys. American Journal of Public Health, 2011, 101(S1): S310‐S317.
[4] Reynolds C, Harris M, Teschke K, Cripton P, Winters M. The impact of transportation infrastructure on
bicycling injuries and crashes: a review of the literature. Environmental Health, 2009, 8(1):47‐65.
[5] Jermakian J. Crash avoidance potential of four passenger vehicle technologies. Accident Analysis and
Prevention, 2011, 43(3):732‐740.
[6] European New Car Assessment Programme. 2020 Roadmap. Brussels, Belgium. 2014.
[7] Kuehn M, Hummel T, Lang A. Cyclist‐car accidents – their consequences for cyclists and typical accident scenarios. Proceedings of the 24th International Conference on the Enhanced Safety of Vehicles, 2015, Gothenburg, Sweden.
[8] Jermakian J, Zuby, D. Primary pedestrian crash scenarios: factors relevant to the design of pedestrian detection systems. Insurance Institute for Highway Safety, Arlington VA, 2011.
[9] Moore M, Zuby D. Collision Avoidance Features: Initial results. Proceedings of the 23rd International Conference on the Enhanced Safety of Vehicles, 2013, Seoul, Republic of Korea.
[10] Fildes B, et al. Effectiveness of low speed autonomous emergency braking in real‐world rear‐end crashes. Accident Analysis & Prevention, 2015, 81:24‐29.
IRC-15-50 IRCOBI Conference 2015
- 384 -